Feasibility of UTE-MRI-based radiomics model for prediction of histopathologic subtype of lung adenocarcinoma: in comparison with CT-based radiomics model

This study evaluates the feasibility of using UTE-MRI-based radiomics to predict micropapillary and/or solid (MP/S) patterns in surgically resected lung adenocarcinoma, compared to a CT-based radiomics model. The study prospectively included 74 lesions from 71 patients who underwent UTE-MRI and CT before surgery. Radiomic features were extracted using Aview Research software for region segmentation and feature extraction. Six models combining conventional radiologic analysis, UTE-MRI Rad-score, and CT Rad-score were constructed and compared using AUC values. The results showed that 24 lesions were MP/S-positive, and 50 were MP/S-negative. Both UTE-MRI and CT Rad-scores had AUCs of 0.84, with the combined model achieving the highest diagnostic performance (AUC = 0.879). Survival analysis indicated that Rad-scores from both UTE-MRI and CT could effectively stratify high- and low-risk groups, correlating with early recurrence. The UTE-MRI radiomic model is feasible and comparable to the CT radiomic model in predicting MP/S positivity and early recurrence.